AI Paper Atlas

Research map / Learning theory and algorithms / Combinatorial and search optimization

Neural Guided Constraint Logic Programming for Program Synthesis

Lisa Zhang, Gregory Rosenblatt, Ethan Fetaya, Renjie Liao, William Byrd, Matthew Might, Raquel Urtasun, Richard Zemel

NeurIPS 2018 · Learning theory and algorithms › Combinatorial and search optimization · approach: Differentiable and symbolic reasoning · 15 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use RNN constraint scoring instead of symbolic miniKanren searchminiKanren · Tree Manipulation ProblemsPercent Solved: 27% -> 99%Improved
use RNN constraint scoring instead of hand-tuned search heuristicsminiKanren · Tree Manipulation ProblemsPercent Solved: 82% -> 99%Improved
use RNN constraint guidance instead of input/output and partial-program embeddings (ablation)neural-guided miniKanren · Tree Manipulation ProblemsPercent Solved: 93% -> 99%Improved
use RNN constraint guidance instead of input/output and partial-program embeddings (ablation)neural-guided miniKanren · Tree Manipulation ProblemsAverage Steps: 46.7 -> 37.0Improved
use GNN rather than RNN to score constraints (ablation)neural-guided miniKanren · Tree Manipulation ProblemsPercent Solved: 99% -> 88%Hurt
use GNN rather than RNN to score constraints (ablation)neural-guided miniKanren · Tree Manipulation ProblemsAverage Steps: 37.0 -> 44.5Hurt
add constraint input to neural search guidance (ablation)RNN-guided synthesis · Repeat(N), DropLast(N), BringToFront(N)Largest N synthesized: 9 (time) -> 20+; 3 (time) -> 6 (time); 2 (time) -> 5 (time)Improved
increase the number of GNN message-passing passes (ablation)GNN-Guided + Constraints · Tree Manipulation ProblemsPercent SolvedNo clear effect
use average-pooling instead of min-pooling (ablation)neural constraint scoring · TrainingGradient flowImproved
expand two candidates per training step instead of one (ablation)neural-guided synthesis · TrainingModel robustnessImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy